📦
Eyeshot CAD
Enables AI-driven CAD operations in applications built on the Eyeshot SDK.
0 installs
Trust: 34 — Low
Science
Ask AI about Eyeshot CAD
Powered by Claude · Grounded in docs
I know everything about Eyeshot CAD. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Loading tools...
Reviews
Documentation
AI-Powered CAD System (MCP + Eyeshot)
This repository contains the hackathon prototype for an AI-driven CAD application utilizing the Model Context Protocol (MCP) and devDept Eyeshot SDK.
Architecture layers:
- frontend: React + Three.js interface combining a Chat Assistant with a 3D Canvas.
- backend-mcp: Python FastAPI server handling the MCP routing logic and tool execution.
- llm-service: Python FastAPI server responsible for talking to the LLM (Gemini/GPT) to map natural language to CAD operations.
- cad-engine: C# ASP.NET Core server utilizing the devDept Eyeshot SDK to perform CAD operations headlessly.
- shared: A set of common schemas mapping commands across the layer divides.
Setup
- Copy
.env.exampleto.env. - Assign your secret keys:
GEMINI_API_KEY: Your Gemini API key from Google AI Studio.EYESHOT_LICENSE_KEY: Your Eyeshot production license key.
- IMPORTANT: NEVER commit your
.envfile to the repository. It is already included in.gitignoreto prevent accidental credential leaks.
Requirements
- Node.js (for the frontend)
- Python 3.10+ (for MCP and LLM services)
- .NET 8 SDK (for the CAD engine)
Command Flow example
- Prompt: "Load the sample.step model"
- MCP router sends to LLM service.
- LLM Service returns:
{"action": "load_model", "file_path": "sample.step"} - MCP router dispatches HTTP POST to CAD Engine with payload.
- CAD Engine executes the command and yields the modified state.